TELKOMNIKATelecommunication,Computing,ElectronicsandControl
Vol.21,No.1,February2023,pp.26∼40
TELKOMNIKATelecommunication,Computing,ElectronicsandControl
Vol.21,No.1,February2023,pp.26∼40
AphirakJansang,ChayathornSimasathien,AnanPhonphoem
IntelligentWirelessNetworkGroup(IWING),DepartmentofComputerEngineering,KasetsartUniversity,Chatuchak,Bangkok10900, Thailand
ArticleInfo
Articlehistory:
ReceivedFeb22,2022 RevisedJul26,2022 AcceptedAug10,2022
Keywords: Energyconsumption Energy-awaredeployment Optimization Routerplacement Wirelessmeshnetwork Wirelessnetworkplanning
Wirelessmeshnetworksarewidelyusedtocreatenetworkinfrastructureinruralareasduetotheirflexibleproperties,suchasself-healingmechanismsandassociated redundantpaths.Forexample,awirelessmeshnetworkcansuitablyoperatewith limitedbatterypowerinwildlifemonitoringapplications.Thelocationsofmobile routerstosupportmobilesensornodesareessentialforextendingthesystemlifetime. Thisstudyproposesanenergy-awarewirelessmeshnetworkdeploymentoptimizationmechanism.Thegoalistodetermineasuitablelocationforthemeshrouters withtheaimofmaximizingnetworklifetime.Afterthelocationsolutionisobtained fromtheproposedmechanism,itisthenevaluatedforsystemlifetimeandnetwork performancebythenetworksimulator(ns-3).Theproposedmethodoutperformsthe brute-forcemethodintermsofcomputationtimeforallamountsofmeshclients.For example,for30meshclients,theproposedmethodusesonlyafewminutes,while thebrute-forcemechanismrequiresmorethan200minutestocompletetheprocess. Furthermore,comparedtothebrute-forcemethod,itachievesnearlythesamesystem lifetimeandotherperformanceparameters,suchasthroughput,packetdeliveryratio, andpacketinter-arrivaltime.Intherealimplementation,inwhichthesensornode placementscanbechangedduringtheinstallationperiodowingtotheenvironmental statusortherecommendationoftheinstallers,theresultscanberecalculatedinashort period.
AnanPhonphoem IntelligentWirelessNetworkGroup(IWING),DepartmentofComputerEngineering,KasetsartUniversity Chatuchak,Bangkok10900,Thailand Email:anan.p@ku.ac.th
Currently,Internetaccessiswidelydeployedinbothurbanandruralareas.Highaccessspeedsin therangeofgigabitpersecondcanbeachievedowingtotheavailabilityoffibretothehome(FTTH)and asymmetricdigitalsubscriberline(ADSL)inurbanareas.Thetechnologiesareforfixedlinelocationandtoo costlyforlongcableinstallationthatarenotsuitableforremotescatteringlocationsinruralareas.Conversely, wirelesscellulartechnologies,suchas3GandLTE,aremorepopularduetotheircoverageareasandmobility support.HighspeedWi-Fiisananothersolutionthatisusedinhouseholdandpublicareas.However,its coverageareaislimitedupto50m–100m.
Forwildlifemonitoringapplications[1],[2],suchastigermonitoringorforestprotection[3],wireless communicationsarerequiredforreal-timeandnearreal-timemonitoring.Inlargewildlifesanctuaryareas,such asHuaiKhaKhaenglocatedinUthaiThaniandTakprovinces,Thailand,servicesofcellulartechnologiesare
Journalhomepage: http://telkomnika.uad.ac.id
onlyavailableattheedgeoftheforest.Unfortunately,atthemonitoringsite,wildlifeanimalsusuallylivein deepforestwherenowirelesssignalisavailableatall.
Toperformwildlifeandforestmonitoring,standalonecameratrapsareinstalledbywildlifepatrollers inthedeepforest.Afterdepletionofbatteriesthatlastforabout30–45days,cameratrapsarebroughtbackto thecampsiteforretrievingcapturedpicturesorvideostoredintheSDcard.Patrollersmustvisiteachsitein harshwalkingconditionsatleasttwice.Furthermore,ifsuspiciousorabnormaleventsoccur,itistooslowfor officerstoadoptadecisionandtakeanyactionthatdoesnotcauseseveredamagetothenaturalresources.
Newversionsofcameratrapsareequippedwithawirelesscommunicationcapabilityfordatauploadingvia3G,LTE,orWi-Fi.However,todeploythecameratraps,wirelessnetworkinfrastructureisrequiredat themonitoringsite.Extendingthecoveragesignalofthecellularnetworkisnotafeasiblesolutionduetoits restrictedinstallationandoperationcostfortheserviceproviders.
Specifically,Wi-Fimeshnetworkexhibitsadhocnetworktopologiesandisaninterestingsolutionto createanetworkinfrastructure.Additionally,Wi-Fimeshnetworkisaself-healingnetworkthatisdesignedto recovercommunicationproblemsduetobrokenlinksormalfunctionednodesbyusinganyalternativeroutes. Inforest,allnodesareevidentlyoperatedwithlimitedbatterypowerthatcanberechargedbyitsinstalledsolar powersystem.However,thesolarchargermaynotalwaysworkduetocloudyskiesortreecanopyconditions. Furthermore,batteriesreplacementrequiressignificanttimeandefforts.Therefore,itisnecessarytominimize energyusagetoprolongnetworksystemuptime.
Tominimizeenergyusage,manyroutingmanagementtechniquesareproposed,includingspectral efficiencyroutingprotocol[4],flowandpowerallocationroutingprotocol[5]–[7],load-awareroutingalgorithm[8],[9],harvestingenergymechanismbasedonenergy-efficientroutingprotocol[10]andanenergyefficientroutingprotocolwithnodemobilitysupport[11].Besides,ahybrid(proactiveandreactive)routing protocol[12]hasalsobeenproposed.Someothermechanismsincludesoftware-definedenergy-awarerouting protocol[13],[14],admissioncontrolforpreventingfastbatterydepletion[15],operatingcertainnodesinsleep mode[16],limitingcertainrouterfunctionsforaperiodofatime[17],[18],energyawaremulti-layerdesign [19]andtopologycontrolmechanism[20],[21].Forchannelmanagement,adynamicchannelassignmentfor wirelessmeshnetwork[22]orimplementingageneticalgorithmandneuralnetworktooptimizeenergy[23] arealsopresented.
Alltheaforementionedenergyawaretechniquesareperformedduringoperationalperiodafternodes areinstalledatthepredefinedlocation.Evidently,locationsofallnodesareamajorfactorforsystemenergy usage.Inappropriatenodelocationscanshortensystemlifetime.Whencertainnodesareidle,somenodes canconstituteasystembottleneckwhereinenergyisquicklydrainedandcannotbealleviatedbyanyrouting adjustmenttechniques.
Itispossibletodetermineoptimallocationsofallrouternodesoff-linebyusingabruteforcetechnique beforetheinstallationperiod.However,thecalculationtimecanbeexcessivelylong.Inarealdeployment situation,locationsofallsensornodesarecarefullyselectedbypatrollers.Duringtheinstallationperiod,the sensornodelocationsnormallyrequiresomeadjustmentsbasedonthesiteenvironmentandanimaltrails. Exactsensornodelocationsmaynotbeknownbefore-hand,thusthemeshrouterlocationfindingsmustbe completedinalimitedperiodduringtheinstallationphase.
Inthestudy,anEnergy-awarewirelessmeshnetworkdeploymentoptimizationisproposed.Themain objectiveistodetermineappropriatelocationsofallmeshroutersinashortperiodoftimebyimplementing optimizationusingtheR-module.Additionally,selectedplacementlocationsshouldmaximizethesystem lifetimeforeffectivelyoperatingandmaintainingthesystemintheforest.Hence,afterobtainingoptimization results,thesystemlifetimeandnetworkperformanceparametersofallsensorsandrouternodesareevaluated andcomparedwithbrute-forcemechanismbyusinganns-3simulator.
Relatedextantstudiesarepresentedinthenextsection.Subsequently,theproposedmechanism,problemformulationandperformanceevaluationaredescribedinsection3,section4,andsection5,respectively. Thesimulationresultsarepresentedinsection6.Finally,section7concludesthestudy.
Forthewirelessmeshnetworkplacement,mostextantstudiesminimizethedeploymentcost.Forexamplein[24],themixed-integerlinearprogrammingmodeltoreducetheinstallationcostofthewirelessmesh networkisproposedbyaccountingfortrafficrouting,interference,rateadaptation,andchannelassignment.
Energy-awarewirelessmeshnetworkdeploymentusingoptimizationmechanism(AphirakJansang)
Furthermore,themechanismcanimprovethecalculationtimewiththesamenetworkperformancebyusinga heuristicalgorithm.However,energyawarenessisnotconsidered.Byextendingthealgorithmffrom[24],an enhancedalgorithm[25]usesthetrafficprofilestoturnsomenodesonandofftominimizeenergyusagewhile therequirednodetransmissionisstillsatisfied.
Later,twomathematicalmodels[26],subsequenceimprovement[24],[25],forenergy-awarewireless meshnetworkmanagementhavebeenproposedtofindthebalancebetweentheinstallationcostandenergy managementoperationcost.Thefirstmodelaimsatlocationfindingsbasedoninstallationcost.Incontrast,the secondmodelseekstominimizeenergybyturningon-offroutersduringtheoperationperiodbasedonnetwork demands.
Nevertheless,forrechargeableMRplacementwithbatterylimited,twoMRassociationmechanisms foreachnodehavebeenproposed[27].Furthermore,theenergycharginganddischargingmodelsarealso includedintheproposedoptimizationmodelstoprotectrouterenergyfrombeingdepleted.Similarly,Wang etal. [28]usedthesameconstraintsintheoptimizationmodel.However,notonlytheMRplacementbut alsotheInternetgatewayhavebeenincorporatedtominimizetheinstallationandenergyconsumptioncostsby usingthetime-divisionmultiplexingslottedmechanismfortraffictransmission.In[24]–[28],allmechanisms focusonfindingtheMRlocation.However,inthecaseoftheinstalledmeshnetwork,addingmoregateway forreducingtheenergyusagehasbeenproposed[29].Themixed-integerlinearprogrammingandagreedy algorithmarepresentedandcompared.Theresultsshowthatthemixed-integermechanismgivestheoptimal result.Meanwhile,only5%gainstheoptimalsolutionforthegreedyalgorithmwithonlyonepercentofthe computationaltime.
Aheuristicmethod[30]isproposedtofindthemaximumnumberofnoninterferinglinksforcreating concurrenttransmissionpathsinafixedinstallednodeenvironmentscenario.Theenergy-savingcomponent wasnotaddressedinthisstudy.Twomixed-integerlinearprogrammingandalocalsearchalgorithmtodeterminetheoptimalgatewaycandidatelocation[31]isproposedwiththeassumptionofunlimitedpowersource formeshrouter.
Thenodeplacementmechanismsbyusingintelligenthybridsystemconsistingofparticleswarmoptimization,hillclimbing,andsimulatedannealing[32]isproposed.Byfixingthenumberofmeshrouters, theroutersareplacedforoptimalcoverageforallclientnodesbasedoncomputationaltimeconcern.However,performanceevaluationisomitted.Anothermeshrouterplacementisproposedbyimplementingthe electromagnetism-likemechanism(EM)metaheuristictechnique[33].Thestudyfocusesonnetworkconnectivityofallclientnodes.Energyharvestismentionedinafuturestudy.
Theoptimizinggoalofmostpreviousresearchinitiallyfocusesonminimizingtheinstallationcost. Theenergyawarenessisthenextinterestingobjective.However,theproposedprotocolalsofocusesonmaximizinglifetimebesidesenergyawareness.RelatedresearchissummarizedasshowninTable1.
Table1.Relatedworkcomparison RelatedworkEnergy-waredesignRechargableenenrgyOptimizationmethodMinimizeinstallationcostMaximizelifetime Amaldi etal.[24]
Capone etal.[25]
Boiardi etal.[26]
Huan etal.[27]
Wang etal.[28]
Sakamoto etal.[32]
Sayad etal.[33]
Thescopeofthestudyistodetermineappropriatemeshrouter(MR)locationsinaforestinashort periodoftime.Allsensornodesaretermedasmeshclient(MC)includingcameratrapsthatareassumedto beplacedintheforestforonlyuploadinginformationtoMR,whichiseventuallyforwardedtothegateway. Thelocationsofsensingdevices,MCs,aredeterminedbythepatrolofficerinarangeoffewkilometers.
TELKOMNIKATelecommunComputElControl,Vol.21,No.1,February2023:26–40
Figure1.Humanhabitatandforestarea
Cellularservicesarenotavailable.Furthermore,sensinginformationcanbevariedbyafewbytesinvolving temperatureandhumidityvaluestocaptureimagesorvideoswithlargefilesize,andtherebynecessitatinghigh datatransmissionrates.Evidently,therequiredscenariocannotbesupportedbytheLoRAtechnologythatis designedforlongrangecommunicationswithlowpowerandaverylimitedtransmissionrate.Thegatewayis locatedattheforestedgenearavillage,campsite,orhumanhabitatwhereInternetaccessandelectricityare available.ThedeploymentscenarioexamplemapisshowninFigure1.
Inthestudy,eachMRisoperatedonlybyastandalonebatterywithoutsolarsystem.Hence,theMR lifetimedependsuponitsoperationandtransmissionactivities.WeassumethatallMRsareinstalledwithsame batterycapacity.Inordertoevaluatethescenario,Wi-Ficommunication(IEEE802.11g,2.4GHz,54Mbps) technologyisselected.
Themonitoringareaisdefinedas M × M meters,whichisdividedintosmallgridcellsbasedonits Wi-Ficommunicationcapabilities.Inthestudy,theplacementscenarioexampleisshowninFigure2(a)where Mequals600m.TheMCcanbeplacedanywhereinthemonitoringareathatisthenmappedintothecenter ofaMCgridcell, 50m × 50m insize,asshowninFigure2(b).Afterperformingtheoptimization,allMR locationscanbefoundandplacedinthecenterofaMRgridcell, 100m × 100m insize,asshowninFigure 2(c).ItisnotedthattodecreasetheMRlocationsearchspaceinourproposedoptimizationmodel,theMR gridcellistwicetheMCgridcell.
ByassumingthatthetransmissionrangeforeachsensingnodetoMRandviceversaissetto250 mfromthecenter,itapproximatelyequalstothenext4MCgridcells.Conversely,thetransmissionrange betweenMRsissetto550m,whichapproximatelycorrespondstothenext5MRgridcells.Hence,with respectto 600m × 600m,fourMRscanextendcoverageinalltheareas.However,theadditionofonemore MRincreasesflexibilityforMRlocationadjustments.
TheconstraintfortheMRplacementisthateachMCmustbecoveredbyatleastoneMRtoensure connectivity.WeassumethatallsensingdevicesareequippedwithanSDcardandareusedtoperiodically uploadinformationataconstantbitrate(CBR).Furthermore,physicaldatatransmissionrateisconsideredas afixedvalueindependentofthecommunicationdistance.
Theproposedmodelisformulatedasamixedintegerlinearprogramming.Inordertodeterminethe optimizedsolution,Rprogramwiththelinprogpackage[34]isused.Fromanetworkviewpoint,thederived locationsfromthemodelareevaluatedwithns-3(networksimulator3)[35]fornetworkperformanceand batteryutilizationthataresubsequentlycomparedwiththebrute-forcemechanism.
Energy-awarewirelessmeshnetworkdeploymentusingoptimizationmechanism(AphirakJansang)
Figure2.Devicemappinginthemonitoringareain(a)sensingdeviceandgatewaylocations,(b)MCgridcell (50m × 50m)and(c)MRgridcell(100m × 100m)
4.1. Datapre-processing
ThelocationsofallMCsandgatewayareusedastheinputoftheproposedmodelthatissubsequently mappedtogridcellsintheoptimizationmodelparameters,asshowninFigure2(a).Let C = {1,...,c}, R = {1,...,r},and G = {1,...,g} bethesetofMC,MR,andgatewaygridcells,respectively.EachMC locationisassignedsuchthatitisamemberofacorrespondingMCgridcell.InMCgridcell i pointof view,thereachableparameter aij ofeachMRgridcell j thatcommunicateswith(fromeachMClocationto reachableMRlocation)each i ∈ C,j ∈ R is: aij = 1, MCgridcell i coveredbyMRgridcell j 0, otherwise (1)
1
0
(2)
Allthereachablecellsfromgatewayarecalculatedandstoredasgatewayrangeparameter wjg j ∈ R,g ∈ G: wjg =
1, ifMRinlocation j andgatewayinlocation g areineachcommunicationrange 0, otherwise
(3)
Next,wedefineadecisionvariable.TheMRassignmentvariable zj foreach j ∈ R: zj = 1, ifMRisinstalledatlocation j 0, otherwise (4)
TheMCassignmentvariable xij foreach i ∈ C,j ∈ R: xij = 1, ifMCatlocationi connectwithMRatlocation j 0, otherwise (5)
TheMRconnectionvariable yjl foreach j,l ∈ R: yjl = 1, ifMRatlocation j connectwithMRatlocation l 0, otherwise (6)
Thegatewayassignmentvariable Yjg foreach j ∈ R,g ∈ G: Yjg = 1, ifMRatlocation j connectwithgatewayatlocation g 0, otherwise (7)
Fromtheobjectivefunction,theoptimallocationforallspecificnumberofMRsthatmakesthelongest operationlifetimeiscalculated.ThesystemlifetimethatisdefinedasoneofMRsisdepleted.Givencertain constraints,thesystemlifetime T ismaximized(orminimizingtheinverseof T )(8).
The xij isabinaryoperatorthatcontrolstheassociationbetweenMRandMC.IfMC i canconnect withMR j,thesumof xij foreachMC i shouldbeatleast1.Therefore,constraints(9)ensuresthatMC i can associatewithonlyoneMR j.Whileconstraints(10)arecoherenceconditionstoguaranteethatMC i canbe associatedwithMR j onlyifthereexistsMRat j iswithincommunicationrangeofMCat i Constraints(11)isdefinedtheflowbalanceatMR j,whichistheinbounddataratethatneedstobe equaltotheoutbounddatarate.Let di beademandeddataratethatMC i tosendtoanyMRs.Let Frlj be thedataratethatMR l requirestosendtoMR j andlet Fgjg bethedataratethatMR j requirestosendto gateway g.Theterm i di xij isthetotalneededdataratetotransmittoassociatedMR j l Frlj isthe totaldataratethatMR i neededtosendtoMR j.ItisthetotalinboundtrafficforMR j.Incontrast, l Frjl and g Fgjg arethetotaltrafficofMR j senttootherMRsandgateways,respectively,whichistheoutbound direction.
Let Capjl bethecapacityofthewirelesslinkbetweenMR j andMR l.Constraints(12)ensurethat thelinkbetweenMR j andMR l existsandthesumoftheoutbounddatarateandinbounddataratebetween MR j andMR l mustnotexceedthelink’scapacity.Constraints(13)–(15)definetheexistenceofawirelesslink betweenMR j andMR l,whichdependsontheexistenceofMRatlocation i,location l andcommunication rangevariable bjl.Constraints(16)ensurethatthenumberoflinksbetweenMRtoMRandMRtogateways doesnotexceedthelink’sdegree.Thus,thesumofanyMR j’slinks(MRtoMR,MRtogateways)mustnot surpassits Degj ,where Degj bethemaximumdegreeofMR j
SincetheconnectionfromanyMRstothegatewayisonlyuplinkdirection.Thus,constraints(17) ensuresthatlinkfromMR j togateway g mustexistandlimitby Capjg,andonlyoutboundtrafficfromMR j tothegateway g isconsidered.Let Capjg bethecapacityofthewirelesslinkbetweenMR j tothegateway g.Constraints(18)definetheexistenceofawirelesslinkbetweenMR j andgateway g,whichdepends
Energy-awarewirelessmeshnetworkdeploymentusingoptimizationmechanism(AphirakJansang)
ISSN:1693-6930
ontheexistenceofMRatlocation i,thegatewayatlocation g,andthecommunicationrangevariable wjg. Constraints(19)ensurethatthetotalnumberoflinksfromanypossibleMRstogateway g mustnotexceedits degree,asdenotedby Degg
Let Enerj betheinitialenergyofMR j.Let T denotethesystemlifetimethatstartsfromthestartup timeuntilthefirstMRrunsoutofbattery.ThenumberofrequiredMRsisdefinedbyvariable numr.Given thatenergyconsumptiononeachMRdependsonitsdatarateandsystemlifetime T .Theoutbounddatarate forMR j issummingupthedataratefromotherMRstoMR j (definedas l Frlj )anditsrequireddatarate togateway g (definedas g Fggj ).Then,theconstraint(20)isusedtoensurethattheenergyusageofeach MR j ( l Frlj + g Fggj multipliedbytime T )willnotexceedits Enerj .Constraints(21)ensurethatthe totalnumberofselectedMRs( j zj )mustnotexceed numr
Fromtheequationshownabove,therearemanypossiblesetsofMRlocations(zj )thatarequalified byallconstrains.However,theobjectiveoftheproposedmechanismistofindforthebestoutcomeof zj whichyieldsthelongestsystemlifetime.Therefore,theoptimizationsolverisrequiredforobtaining zj ,which deliversthebestresult.
Maximize: T (8)
Subjectto: j
xij =1, ∀i ∈ C (9)
xij ≤ aij · zj , ∀i ∈ C, ∀j ∈ R (10)
i di xij + l Frlj = l Frjl + g Fgjg, ∀j ∈ R (11)
Frjl + Frlj ≤ Capjl yjl, ∀j,l ∈ R (12) yjl ≤ bjl, ∀j,l ∈ R (13)
yjl ≤ zj , ∀j,l ∈ R (14) yjl ≤ zl, ∀j,l ∈ R (15)
l yjl + g
Yjg ≤ Degj , ∀j ∈ R (16)
Fgjg ≤ Capjg Yjg, ∀g ∈ G, ∀j ∈ R (17)
Yjg ≤ wjg, ∀g ∈ G, ∀j ∈ R (18) j Yjg ≤ Degg, ∀g ∈ G (19)
( l Frlj + g Fggj ) T ≤ Enerj , ∀j ∈ R (20) j zj ≤ numr, ∀j ∈ R (21)
Inthestudy,fivecasesofvariousMCnodes(i.e.15,20,25,30,and45)areevaluated.Ineachcase, threereplicationsoftheMClocationsarerandomlygeneratedwithuniformdistribution.Thegatewaylocation isassumedasfixedforallcases.Forsystemlifetimeevaluation,eachbrute-forcesolution(fromallpossible MRlocations)foreachcaseissimulatedandcomparedwiththecorrespondingoptimizationsolution.
BeforetheproposedoptimizationmodelcandeterminetheresultsofpotentialMRlocationsonthe map,thebasicinformationfromtheactualmap,suchastheMCs’locationsandthelocationofthegateway usedtorelaydata,mustfirstbegathered.Thecollecteddatawillthenbeconvertedintothemodel’sinitial vectorvariables,suchasmatrix aij forMClocations, wjg forgatewaylocations,whichisdescribedinsection 4.1.Next,initializethemodel’svariablesbycalculatingallpossibleMRlocationpairsinthedeployment areafromvector bjl andgeneratingthemintothegridcellasmatrix wjg.Atthesametime,thematrix Yjg isgeneratedfromvector wjg tokeepallpossibleMRlocationsthatcancommunicatewithgatewayg.The otherrequiredvectorvariables(e.g. zj , xij ,and ylj )andallconstraintsaresetupbeforerunningthelinear optimizationsolver.Subsequently,theprogramforsolvinglinearequationswillthenstarttosearchforthe possiblesolutionsthatmakethemostextendedsystemlifetimewhilesatisfyingallconstraints.Thesoftwareis writteninRandsolveslinearprogrammingtaskswiththe linprog package.Afterobtainingtheoptimization resultfortheMR’slocationsingridcellformat,itwillconvertthembacktotheiractualmaplocationsbefore beingusedintherealimplementation.TheflowchartinFigure3illustratestheprocessofextractingtheresults fromtheproposedoptimizationmodel.
Start
Collect required information on the map (e.g. locations of gateway and MCs)
OptimizationModel
Convert locations into model's grid cells
Initial model's variables and constrained
Optimize MR's locations with linprog
Collect All MR's grid cells locations which is satisfed constrained
Convert All MR's grid cells into map's locations
End
Figure3.Theprocessofextractingtheresultsfromtheproposedoptimizationmodel
Energy-awarewirelessmeshnetworkdeploymentusingoptimizationmechanism(AphirakJansang)
Tocomparethenetworkperformancebetweentheproposedandbrute-forcemechanisms,theoptimizationresultsfromtheR-programaresimulatedbyns-3simulator.Thesimulationfieldareaisdefinedas 600 × 600 m2.TheMCgridcellaredefinedas 50 × 50 m2 forplacingeachMC.Conversely,theMRgrid cellof 100 × 100 m2 isusedforMRplacement.TheMCtrafficpatternsforuploadingdatatothegateway aredefinedataconstantbitrateof1Mbps.AllMRsaredeployedwiththesameinitialcapacitiesandenergy dischargemodel.Specifically,IEEE802.11gwithfixedpropagationdelaywiththedefaultrangepropagation lossmodelisusedthroughoutthesimulation.Wemodifiedtheoptimizedlinkstateroutingprotocol(OLSR)to supportanenergy-awareroutingprotocolbyconsideringtheremainingenergyineachpathbeforetransmitting thedata.Thetotalsimulationtimecorrespondsto80s.EachMCautomaticallystartsuploadingatthe40th secondafterthesimulationisstarted.ThesimulationparametersarelistedinTable2.
Table2.Parameterofsimulation ParametersValue
Fieldsize 600 × 600 m2
Client’sgridsize 50 × 50 m2 Router’sgridsize 100 × 100 m2
Communicationrange(router-client)250m
Communicationrange(router-router)550m Client’sapplicationCBRat1MbpsperMC
Energymodelns-3energymodel
Initialenergy80J WirelessstandardIEEE802.11g/54Mbps DelaymodelConstantspeedpropagation LossmodelRangepropagation RoutingprotocolOLSRwithenergy-awaresupport Totalsimulationtime80s
Applicationstarttime40ths
Numberofclientspercase15/20/25/30/45 Numberofexperimentspercase3 Numberofrouter5 SearchingmethodOptimization/brute-force
Thedefaultns-3energymodelisusedthroughoutthesimulation.Theenergyusagesarebasedon itsactivities,namelytransmit,receive,oridle,asshowninTable3.EachMRstopsworkingifitrunsoutof energy.
Table3.Energymodelusedinns3[35]
ActivityCurrentusage(ampere) Idle0.273 Receive0.313 Transmit1.800
Theproposedmechanismisevaluatedforthecomputationtimeandnetworkperformancebycomparingitwiththebrute-forcemethod.AsshowninFigure4,thecomputationaltimes(includingthesimulation times)forvariousMCsoftheproposedoptimizationarealmostconstant.Furthermore,thebrute-forcemechanismcomputationtimeissignificantlyvariedandsignificantlyhighwhencomparedtotheproposedmechanism.Withrespecttonetworkperformance,thesystemlifetimes,asshowninFigure5,forbothmechanisms arealmostidentical,andtheybothrunoutofbatteryatapproximately65thsecond.Fortheaveragethroughput andpacketdeliveryratio,showninFigures6and7,forthelownumberofclients,thepredefinedcondition oftheoptimizationmightrestricttheallowabledegreeofMCconnectingforeachMR.Therefore,oncethe increasingnumberofclientsreaches30,bothbrute-forceandoptimizationmechanismsperformthesamedue tothelimitedchoicesofMRlocationforcoveringallMCs.However,inFigure8,theaveragedelayvariation
TELKOMNIKATelecommunComputElControl,Vol.21,No.1,February2023:26–40
ofthebrute-forcemechanismformanyconnectingclientsshowsasignificantvariationduetotheunbalanced connectingdegreeforeachMR.
FromFigure9,anexampleof25clientsrandomlyscatteredonthemapispresented.Theresultsof thedifferentmethodsfordeterminingtheplacementof5MRlocationsareshowninFigure10.Theoptimizationsolution(showninFigure10(a)andFigure10(b))yieldsmorescatterMRlocationsthanthebrute-force mechanism(showninFigure10(c)andFigure10(d))becausetheoptimizationmechanismtriedtolimitthe degreeofconnectingMCforeachMR,whichismoreapparentinthecaseofthehighernumberofrequired MRs.
Figure4.Comparisonofcomputationtimebetweenmethods
Figure5.Comparisonoflifetimebetweenmethods
Energy-awarewirelessmeshnetworkdeploymentusingoptimizationmechanism(AphirakJansang)
Figure6.Comparisonofaveragethroughputbetweenmethods
Figure7.Comparisonofpacketdeliveryratiobetweenmethods
Figure8.Comparisonofaveragedelaybetweenmethods
TELKOMNIKATelecommunComputElControl,Vol.21,No.1,February2023:26–40
Figure9.Clientlocations
Figure10.ComparisonofMRplacementbetweenbrute-forceandoptimizationmethods(for25clients)in(a) clientconnection:brute-forcemethod,(b)routerconnection:brute-forcemethod,(c)clientconnection: optimizationmethod,and(d)routerconnection:optimizationmethod Energy-awarewirelessmeshnetworkdeploymentusingoptimizationmechanism(AphirakJansang)
Theinter-arrivaltimeofthecaseof25MCs(fromFigure10)isshowninFigure11.Giventhe cumulativedistributionfunction(CDF)valueof0.80,theperformanceoftheoptimizationmethodsexceeds thatofthebrute-forcemethod.Additionally,withrespecttodifferentnumberofMCnodes,theaverageof inter-arrivaltimeat80thpercentileisshowninFigure12.Theperformanceoftheproposedmethodslightly exceedsthatofthebrute-forcemethodforahighnumberofMCnodes.Whencomparedtothebrute-force mechanism,thenetworkperformanceobtainedbyourproposedmechanismaresignificantlyidenticalwhile thecomputationtimesignificantlyoutperformsthebrute-forcemechanism. 0.8 CDF
Figure11.Inter-arrivaltimeof25clients
Figure12.Theaverageofinter-arrivaltimeat80thpercentile
Inthestudy,weproposedtheoptimizedwirelessrouterplacementmethodwithenergyawarefor wirelessmeshnetworkinaruralarea.Ouroptimizationgoalinvolvesachievingthelongestlifetimeofthe systemcomparedwithbrute-forcemethod.Thesimulationresultsindicatethattheoptimizationmethodyields significantlylesscomputationtimewhileobtainingthesameresultsintermofnetworkparameters.This issignificantlycrucialintherealdeploymentsituationwheresensornode(client)locationsaremandatory adjustedbasedontheenvironmentalstatusortherecommendationofthepatrollers.Eachadjustmentrequires
anewcalculation.Therefore,theproposedmechanismbecomesmoredesirableduetoitsshortcomputation time.Furthermore,theheuristicapproachcanbeimplementedforbetterperformance.
[1] A.M-BadescuandL.Cotofana,“Awirelesssensornetworktomonitorandprotecttigersinthewild,” EcologicalIndicators,vol. 57,pp.447-451,Oct.2015,doi:10.1016/j.ecolind.2015.05.022.
[2] H.J.Griebling,C.M.Sluka,L.A.Stanton,L.B.Barrett,J.P.BastosandS.B-Amran,“Howtechnologycanadvancethestudyof animalcognitioninthewild,” CurrentOpinioninBehavioralSciences,vol.45,Jun.2022,doi:10.1016/j.cobeha.2022.101120.
[3] “EmergingTechnologies:Smarterwaystofightwildlifecrime,” JEnvironmentalDevelopment,vol.12,pp.62-72,Oct.2014,doi: 10.1016/j.envdev.2014.07.002.
[4] L.Zhou,G.Kang,N.Zhang,andJ.Cheng,“Spectralefficiencyguaranteedsustainableroutingforenergyrenewablewirelessmeshnetworks,” 2015InternationalConferenceonWirelessCommunicationsSignalProcessing(WCSP),Oct.2015,doi: 10.1109/WCSP.2015.7341063.
[5] C.Luo,S.Guo,S.Guo,L.T.Yang,G.Min,andX.Xie,“GreenCommunicationinEnergyRenewableWirelessMeshNetworks: Routing,RateControl,andPowerAllocation,” IEEETransactionsonParallelandDistributedSystems,vol.25,no.12,pp.32113220,Dec.2014,doi:10.1109/TPDS.2013.2297922.
[6] A.S.ArezoomandandM.Pourmina,“Prolongingnetworkoperationlifetimewithnewmaximumbatterycapacityroutinginwirelessmeshnetwork,” 2010The2ndInternationalConferenceonComputerandAutomationEngineering(ICCAE),Feb.2010,pp. 319-323,doi:10.1109/ICCAE.2010.5451679.
[7] S.AvalloneandA.Banchs,“AChannelAssignmentandRoutingAlgorithmforEnergyHarvestingMultiradioWireless MeshNetworks,” IEEEJournalonSelectedAreasinCommunications,vol.34,no.5,pp.1463-1476,May2016,doi: 10.1109/JSAC.2016.2520238.
[8] Y.Chai,W.Shi,andT.Shi,“Load-awarecooperativehybridroutingprotocolinhybridwirelessmeshnetworks,” AEU-International JournalofElectronicsandCommunications,vol.74,pp.135-144,Apr.2017,doi:10.1016/j.aeue.2017.02.002.
[9] N.A.Macabale etal., “CRADLE:Cross-LayerDesignforLoad-AwareRoutinginIEEE802.11-basedWirelessMeshandSensor Networks,” 202010thAnnualComputingandCommunicationWorkshopandConference(CCWC),Jan.2020,pp.0970-0974,doi: 10.1109/CCWC47524.2020.9031245.
[10] Y.Yu,Y.Peng,Y.Liu,L.Guo,andM.Song,“SurvivableRoutingProtocolforgreenwirelessmeshnetworksbasedonenergy efficiency,” ChinaCommunications,vol.11,no.8,pp.117-124,Aug.2014,doi:10.1109/CC.2014.6911093.
[11] R.AlmesaeedandA.Jedidi,“Dynamicdirectionalroutingformobilewirelesssensornetworks,” AdHocNetworks,vol.110,Jan. 2021,doi:10.1016/j.adhoc.2020.102301.
[12] Y.Chai,W.Shi,T.Shi,andX.Yang,“Anefficientcooperativehybridroutingprotocolforhybridwirelessmeshnetworks,” Wireless Networks,vol.23,pp.1387–1399,2017.[Online].Available:https://link.springer.com/article/10.1007/s11276-016-1229-8
[13] H.Lin,J.Hu,L.Xu,Y.Tian,L.Liu,andStewartBlakeway“Atrustworthyandenergy-awareroutingprotocolin software-definedwirelessmeshnetworks,” ComputersElectricalEngineering,vol.64,pp.407-419,Nov.2017,doi: 10.1016/j.compeleceng.2016.10.015.
[14] Y.ChaiandX-J.Zeng“LoadBalancingRoutingforWirelessMeshNetworkWithEnergyHarvesting,” IEEECommunications Letters,vol.24,no.4,pp.926-930,Apr.2020,doi:10.1109/LCOMM.2020.2969194.
[15] L.X.Cai,Y.Liu,T.H.Luan,X.S.Shen,J.W.Mark,andH.V.Poor,“SustainabilityAnalysisandResourceManagementfor WirelessMeshNetworkswithRenewableEnergySupplies,” IEEEJournalonSelectedAreasinCommunications,vol.32,no.2, pp.345-355,Feb.2014,doi:10.1109/JSAC.2014.141214.
[16] S.ChenandG.M.Muntean,“E-Mesh:Anenergy-efficientcross-layersolutionforvideodeliveryinwirelessmeshnetworks,” IEEE internationalSymposiumonBroadbandMultimediaSystemsandBroadcasting,2012,pp.1-7,doi:10.1109/BMSB.2012.6264252.
[17] A.Gladisch,R.Daher,P.Lehsten,andD.Tavagarian,“Context-awareenergymanagementforenergy-self-sufficientnetworknodes inwirelessmeshnetworks,” 20113rdInternationalCongressonUltraModernTelecommunicationsandControlSystemsand Workshops(ICUMT),Oct.2011,pp.1-8.[Online].Available:https://ieeexplore.ieee.org/document/6078992
[18] S.Mamechaoui,F.Didi,S.M.Senouci,andG.Pujolle,“Energy-awaredesignforwirelessmeshnetworks,” 2014GlobalInformation InfrastructureandNetworkingSymposium(GIIS),Sep.2014,pp.1-6,doi:10.1109/GIIS.2014.6934259.
[19] J.Hu,L.L.Yang,andL.Hanzo,“Energy-EfficientCross-LayerDesignofWirelessMeshNetworksforContentSharing inOnlineSocialNetworks,” IEEETransactionsonVehicularTechnology,vol.66,no.9,pp.8495-8509,Sep.2017,doi: 10.1109/TVT.2017.2678167.
[20] A.V-RodasandL.J.D.L.C.Llopis,“Acentrality-basedtopologycontrolprotocolforwirelessmeshnetworks,” AdHocNetworks, vol.24,pp.34-54,Jan.2015,doi:10.1016/j.adhoc.2014.07.026.
[21] I.L-Cherif,L.Zitoune,andV.Veque,“EnergyEfficientRoutingforWirelessMeshNetworkswithDirectionalAntennas:When Q-learningmeetsAntsystems,” AdHocNetworks,vol.121,Jul.Oct.2021,doi:10.1016/j.adhoc.2021.102589.
[22] N.Zlobinsky,D.L.Johnson,A.K.Mishra,andA.A.Lysko,“ComparisonofMetaheuristicAlgorithmsforInterface-Constrained ChannelAssignmentinaHybridDynamicSpectrumAccess–Wi-FiInfrastructureWMN,” IEEEAccess,vol.10,pp.26654-26680, Feb.2022,doi:10.1109/ACCESS.2022.3155642.
[23] B.Prakash,S.Jayashri,andT.S.Karthik,“Ahybridgeneticartificialneuralnetwork(G-ANN)algorithmforoptimizationofenergy componentinawirelessmeshnetworktowardgreencomputing,” SoftComputing,vol.23,pp.2789–2798,Jan.2019.[Online]. Available:https://link.springer.com/article/10.1007/s00500-019-03789-8
[24] E.Amaldi,A.Capone,M.Cesana,I.Filippini,andF.Malucelli,“Optimizationmodelsandmethodsforplanningwirelessmesh networks,” ComputerNetworks, vol.52,no.11,pp.2159-2171,Aug.2008,doi:10.1016/j.comnet.2008.02.020.
[25] A.Capone,F.Malandra,andB.Sanso,“EnergySavingsinWirelessMeshNetworksinaTime-VariableContext,” MobileNetworks andApplications,vol.17,pp.298–311,Apr.2022.[Online].Available:https://link.springer.com/article/10.1007/s11036-011-0339x
[26] S.Boiardi,A.Capone,andB.Sanso,“Jointdesignandmanagementofenergy-awareMeshNetworks,” AdHocNetworks,vol.10, no.7,pp.1482-1496,Sep.2012,doi:10.1016/j.adhoc.2012.04.005.
[27] X.Huan,B.Wang,Y.Mo,andL.T.Yang,“Rechargeablerouterplacementbasedonefficiencyandfairnessingreenwirelessmesh networks,” ComputerNetworks,vol.78,pp.83-94,Feb.2015,doi:10.1016/j.comnet.2014.10.035.
[28] B.Wang,X.Huan,L.T.Yang,andY.Mo,“HybridPlacementofInternetGatewaysandRechargeableRouterswithGuaranteed QoSforGreenWirelessMeshNetworks,” MobileNetworksandApplications,vol.20,pp.543–555,Apr.2015.[Online].Available: https://link.springer.com/article/10.1007/s11036-015-0607-2
[29] U.Ashraf,“Energy-AwareGatewayPlacementinGreenWirelessMeshNetworks,” IEEECommunicationsLetters,vol.21,no.1, pp.156-159,2016,doi:10.1109/LCOMM.2016.2618378.
[30] K.GokbayrakandE.A.Yildirim,“Exactandheuristicapproachesbasedonnoninterferingtransmissionsforjointgatewayselection, timeslotallocation,routingandpowercontrolforwirelessmeshnetworks,” ComputersandOperationsResearch,vol.81,pp.102118,May2017,doi:10.1016/j.cor.2016.09.021.
[31] K.Gokbayrak,“Robustgatewayplacementinwirelessmeshnetworks,” ComputersandOperationsResearch,vol.97,pp.84-95, Sep.2018,doi:10.1016/j.cor.2018.04.018.
[32] S.Sakamoto,K.Ozera,M.Ikeda,andL.Barolli,“ImplementationofIntelligentHybridSystemsforNodePlacementProblemin WMNsConsideringParticleSwarmOptimization,HillClimbingandSimulatedAnnealing,” MobileNetworksandApplications, vol.23,pp.27–33,2018,doi:10.1007/s11036-017-0897-7.
[33] L.Sayad,L.B.Medjkoune,andD.Aissani,“AnElectromagnetism-likemechanismalgorithmfortherouter nodeplacementinwirelessmeshnetworks,” SoftComputing,vol.23,pp.4407–4419,2019.[Online].Available: https://link.springer.com/article/10.1007/s00500-018-3096-y
[34] A.Henningsen,“linprog:Linearprogramming/optimization,”Mar.2022.[Online].Available:https://cran.rproject.org/web/packages/linprog/linprog.pdf
[35] G.F.RileyandT.R.Henderson,“Thens-3networksimulator” Modelingandtoolsfornetworksimulation,pp.15-34,2010,doi: 10.1007/978-3-642-12331-3-2.
AphirakJansang receivedhisB.Eng.,M.Eng.andD.Eng.degreesinComputerEngineeringfromKasetsartUniversity,Bangkok,Thailand,in2000,2003and2012,respectively.Heis currentlyanAssistantProfessorattheDepartmentofComputerEngineering,KasetsartUniversity, Thailand.HeisalsoateamleaderofIntelligentWirelessNetworkGroup(IWING).Hisresearch interestsincludewirelessnetworks,resourcemanagementandqualityofservicesupport.Hecanbe contactedatemail:aphirak.j@ku.ac.th.
ChayathornSimasathien receivedtheB.Eng.andM.Eng.degreeinComputerEngineeringfromKasetsartUniversity,Bangkok,Thailand,in2014and2020respectively.Heis alsoateammemberofIntelligentWirelessNetworkGroup(IWING).Heiscurrentlyworkingat KASIKORNBusiness-TechnologyGroup(KBTG),THAILAND.Hecanbecontactedatemail: chayathorn.s@ku.th.
AnanPhonphoem
receivedhisPh.D.inElectricalandComputerEngineering(2000) fromUniversityofMassachusettsAmherst,M.S.inComputerEngineering(1996)fromUniversity ofSouthernCalifornia(USC),andB.Eng.inElectricalEngineering(1990)fromPrinceofSongkla University,Thailand.HeiscurrentlyanAssociateProfessorandthedirectoroftheIntelligentWirelessNetworkGroup(IWING)atComputerEngineeringDepartment,KasetsartUniversity,Thailand. Hisresearchinterestsincludewirelessnetworks,adhocnetworks,performanceevaluation,andprotocoldesignandanalysis.Hecanbecontactedatemail:anan.p@ku.ac.th.